Journal of Ecology and Rural Environment ›› 2020, Vol. 36 ›› Issue (11): 1485-1494.doi: 10.19741/j.issn.1673-4831.2019.0764

Previous Articles     Next Articles

Extracting Large-scale Pig Farms in Plain River Network Area From GF-2 Image

CHEN Jun-song1,2, SHI Fang1, DU Wei2, LI Wen-jing2, FAN Jia-hui1,2, TIAN Jia-rong1,2, LIAO Xiao-wen3, GUO Peng3, YAO Shi-hao3, LI Ming-shi1, WANG Wen-lin2   

  1. 1. College of Forestry, Nanjing Forestry University, Nanjing 210037, China;
    2. Nanjing Institute of Environmental Sciences, Ministry of Ecology and Environment, Nanjing 210042, China;
    3. School of Geographic Science, Nantong University, Nantong 226007, China
  • Received:2019-09-25 Online:2020-11-25 Published:2020-11-18

Abstract: Accurate control and treatment of pig breeding industry pollution requires relevant departments to accurately and quickly grasp the spatial distribution information of large-scale pig farms, which can be realized by remote sensing technology. Taking the GF-2 image as data source, the large-scale pig farms located in Shizhuang Township and Jiangan Township, Rugao City, Jiangsu Province, in the plain river network area of the middle and lower reaches of the Yangtze River were taken as sample and verification objects, respectively. Based on the sample objects, the compositional elements of them were firstly defined as pigsties, fecal sewage pools and impounding reservoirs. Then, based on the chessboard segmentation, the building area index (BAI) was used to remove the interference of factories and other buildings in cluster on pigsty extraction and the multiresolution segmentation was then implemented. Afterwards, the spectral, geometric and textural features of the above-mentioned ground objects were selected to construct the extraction rule set of large-scale pig farms. Following the same extraction rule set in conjunction with GIS spatial overlay and distance analyses, the above-mentioned three kinds of components of the verification objects were extracted. Ultimately, the spatial consistency of the extracted results was verified by using field investigations. The results show that: (1) for the plain river network area, based on the chessboard segmentation with the segmentation object size of 516, BAI >-0.150 923 could be used to remove the interference of factories and other buildings in cluster on pigsty extraction; (2) after removing the interference, the proposed object-oriented extraction method could effectively extract the spatial distribution of large-scale pig farms, with an overall spatial consistency reaching to 82.24%. In conclusion, using very high spatial resolution remote sensing imagery and removing factories and other buildings in cluster, the method of object-oriented extraction is an effective strategy to extract large-scale pig farms in the plain river network area.

Key words: GF-2 image, chessboard segmentation, multiresolution segmentation, object-oriented extraction, large-scale pig farms

CLC Number: